//! General conversion functions and utilities. mod stft; pub mod window; pub use stft::StftHelper; pub const MINUS_INFINITY_DB: f32 = -100.0; pub const MINUS_INFINITY_GAIN: f32 = 1e-5; // 10f32.powf(MINUS_INFINITY_DB / 20) pub const NOTES: [&str; 12] = [ "C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B", ]; /// Temporarily allow allocations within `func` if NIH-plug was configured with the /// `assert_process_allocs` feature. #[cfg(all(debug_assertions, feature = "assert_process_allocs"))] pub fn permit_alloc T>(func: F) -> T { assert_no_alloc::permit_alloc(func) } /// Temporarily allow allocations within `func` if NIH-plug was configured with the /// `assert_process_allocs` feature. #[cfg(not(all(debug_assertions, feature = "assert_process_allocs")))] pub fn permit_alloc T>(func: F) -> T { func() } /// Convert decibels to a voltage gain ratio, treating anything below -100 dB as minus infinity. #[inline] pub fn db_to_gain(dbs: f32) -> f32 { if dbs > MINUS_INFINITY_DB { 10.0f32.powf(dbs * 0.05) } else { 0.0 } } /// Convert a voltage gain ratio to decibels. Gain ratios that aren't positive will be treated as /// [`MINUS_INFINITY_DB`]. #[inline] pub fn gain_to_db(gain: f32) -> f32 { f32::max(gain, MINUS_INFINITY_GAIN).log10() * 20.0 } /// An approximation of [`db_to_gain()`] using `exp()`. Does not treat values below /// [`MINUS_INFINITY_DB`] as 0.0 gain to avoid branching. As a result this function will thus also /// never return 0.0 for normal input values. Will run faster on most architectures, but the result /// may be slightly different. #[inline] pub fn db_to_gain_fast(dbs: f32) -> f32 { const CONVERSION_FACTOR: f32 = std::f32::consts::LN_10 / 20.0; (dbs * CONVERSION_FACTOR).exp() } /// [`db_to_gain_fast()`], but this version does truncate values below [`MINUS_INFINITY_DB`] to 0.0. /// Bikeshedding over a better name is welcome. #[inline] pub fn db_to_gain_fast_branching(dbs: f32) -> f32 { if dbs > MINUS_INFINITY_DB { db_to_gain_fast(dbs) } else { 0.0 } } /// An approximation of [`gain_to_db()`] using `ln()`. Will run faster on most architectures, but /// the result may be slightly different. #[inline] pub fn gain_to_db_fast(gain: f32) -> f32 { const CONVERSION_FACTOR: f32 = std::f32::consts::LOG10_E * 20.0; f32::max(gain, MINUS_INFINITY_GAIN).ln() * CONVERSION_FACTOR } /// [`db_to_gain_fast()`], but the minimum gain value is set to [`f32::EPSILON`]instead of /// [`MINUS_INFINITY_GAIN`]. Useful in conjunction with [`db_to_gain_fast()`]. #[inline] pub fn gain_to_db_fast_epsilon(gain: f32) -> f32 { const CONVERSION_FACTOR: f32 = std::f32::consts::LOG10_E * 20.0; f32::max(gain, MINUS_INFINITY_GAIN).ln() * CONVERSION_FACTOR } /// Convert a MIDI note ID to a frequency at A4 = 440 Hz equal temperament and middle C = note 60 = /// C4. #[inline] pub fn midi_note_to_freq(note: u8) -> f32 { f32_midi_note_to_freq(note as f32) } /// The same as [`midi_note_to_freq()`], but for arbitrary note numbers including those outside of /// the MIDI range. This also supports fractional note numbers, which is useful when working with /// cents. #[inline] pub fn f32_midi_note_to_freq(note: f32) -> f32 { 2.0f32.powf((note - 69.0) / 12.0) * 440.0 } /// The inverse of [`f32_midi_note_to_freq()`]. This returns a fractional note number. Round to a /// whole number, subtract that from the result, and multiply the fractional part by 100 to get the /// number of cents. #[inline] pub fn freq_to_midi_note(freq: f32) -> f32 { ((freq / 440.0).log2() * 12.0) + 69.0 } #[cfg(test)] mod tests { mod db_gain_conversion { use super::super::*; #[test] fn test_db_to_gain_positive() { assert_eq!(db_to_gain(3.0), 1.4125376); } #[test] fn test_db_to_gain_negative() { assert_eq!(db_to_gain(-3.0), 1.4125376f32.recip()); } #[test] fn test_db_to_gain_minus_infinity() { assert_eq!(db_to_gain(-100.0), 0.0); } #[test] fn test_gain_to_db_positive() { assert_eq!(gain_to_db(4.0), 12.041201); } #[test] fn test_gain_to_db_negative() { assert_eq!(gain_to_db(0.25), -12.041201); } #[test] fn test_gain_to_db_minus_infinity_zero() { assert_eq!(gain_to_db(0.0), MINUS_INFINITY_DB); } #[test] fn test_gain_to_db_minus_infinity_negative() { assert_eq!(gain_to_db(-2.0), MINUS_INFINITY_DB); } } mod fast_db_gain_conversion { use super::super::*; #[test] fn test_db_to_gain_positive() { approx::assert_relative_eq!( db_to_gain(3.0), db_to_gain_fast_branching(3.0), epsilon = 1e-7 ); } #[test] fn test_db_to_gain_negative() { approx::assert_relative_eq!( db_to_gain(-3.0), db_to_gain_fast_branching(-3.0), epsilon = 1e-7 ); } #[test] fn test_db_to_gain_minus_infinity() { approx::assert_relative_eq!( db_to_gain(-100.0), db_to_gain_fast_branching(-100.0), epsilon = 1e-7 ); } #[test] fn test_gain_to_db_positive() { approx::assert_relative_eq!(gain_to_db(4.0), gain_to_db_fast(4.0), epsilon = 1e-7); } #[test] fn test_gain_to_db_negative() { approx::assert_relative_eq!(gain_to_db(0.25), gain_to_db_fast(0.25), epsilon = 1e-7); } #[test] fn test_gain_to_db_minus_infinity_zero() { approx::assert_relative_eq!(gain_to_db(0.0), gain_to_db_fast(0.0), epsilon = 1e-7); } #[test] fn test_gain_to_db_minus_infinity_negative() { approx::assert_relative_eq!(gain_to_db(-2.0), gain_to_db_fast(-2.0), epsilon = 1e-7); } } }